We propose a new analytical approach based on ordinal pattern analysis to investigate respiratory heart rate variability (RespHRV, also called respiratory sinus arrhythmia), specifically modulation of heart rate across different phases of the respiratory cycle. The method uses RR interval time series derived from ECG signals, along with simultaneous respiratory recordings obtained with a respiratory belt. The method produces distributions of ordinal patterns that reflect the dynamics of heart rate variability throughout the respiratory cycle. We systematically test how variations in parameters defining ordinal patterns affect the results and interpretation, and discuss the optimal parameter configuration for quantification of RespHRV in short-term recordings. Finally, we demonstrate the ability of the method to differentiate between healthy controls and patients with obstructive sleep apnea based on daytime cardiorespiratory data.

In a scientific and implementation consortium, we developed an adaptive AI platform that enables doctors to create accurate and comprehensive Electronic Health Records (EHRs) through advanced speech recognition and context analysis tailored to Polish medical language. This system ensures stability with consistent performance across real-world clinical settings, achieving expected values for speech and context recognition during extensive testing. Its robustness is demonstrated by handling diverse inputs—such as regional accents, complex terminology, and noisy environments—supported by error-correction mechanisms and a specialized acoustic probe. Sustainability is achieved through seamless integration with existing healthcare infrastructures, scalable design, and ongoing updates to medical dictionaries, facilitating long-term use and adaptation. Structured data from electronic health records (EHRs) supports scientific research based on Real-World Data (RWD), verified by medical specialists using evidence-based medicine (EBM). The platform covers 10 clinical situations. The applied method was illustrated using one situation—a breast X-ray examination—employing clinically approved structures and real-world validation. Approved by the Bioethics Committee, the system is currently being tested at the hospital, marking a significant step toward efficient, reliable, and sustainable healthcare documentation.

The Makowiec Generalised Asymmetry Index, denoted by MI (q) , is a q-parameterised family of directional asymmetry measures for time series, computed from short-term recordings to quantify the imbalance between opposite changes (increases vs. decreases) across fluctuation magnitudes. In this study, we apply the Makowiec Generalised Asymmetry Index to daytime recordings of heart rate variability (HRV) and blood pressure variability (BPV) collected during wakefulness, and we evaluate its usefulness for the characterization of healthy men and obstructive sleep apnea (OSA) patients. Our findings suggest that asymmetry quantified by the Makowiec Generalised Asymmetry Index in daytime for blood pressure signals during wakefulness may serve as a noninvasive marker supporting OSA screening and risk stratification, motivating further validation in larger prospective cohorts.

An accurate, well-organised report from a radiological examination, in our case a breast ultrasound, is an important part of a radiologist's work [1]. As part of the ADMEDVOICE research consortium [6], we have developed an AI platform adapted to Polish medical language, which generates structured reports based on freely dictated text during or after a breast ultrasound examination. We developed a report structure, on the basis of which we created a contextual division of information and a dictionary of synoptic data. Then, using data from real radiology reports, we trained a language model to recognise speech and correctly classify descriptions. We added algorithms for numerical and dictionary data to the grouped information and then put everything into an application that is convenient for doctors to use. Focusing on Polish breast radiology reports, we used BERTopic for thematic modelling to categorise the text. We found that the topics extracted by the model were closely aligned with the categories defined by clinical experts. However, the accuracy of the solution was not satisfactory, reaching 73%. We then tested the creation of a classification using a large language model (LLM) as an information classifier. The LLM model was used as a learning model during the training of a smaller encoder-type language model for text classification. We used two LLM models, the multilingual Llama3.1-70B-Instruct and Bielik-11B-v2.3-Instruct, achieving a classification accuracy of 94%. In the next step, to improve the results, we added appropriate dictionaries of medical terms, abbreviations, and units of measurement used in the descriptions.

In the medical field, text annotation involves categorizing clinical and biomedical texts with specific medical categories, enhancing the organization and interpretation of large volumes of unstructured data. This process is crucial for developing tools such as speech recognition systems, which help medical professionals reduce their paperwork. It addresses a significant cause of burnout reported by up to 60% of medical staff. However, annotating medical texts in languages other than English poses unique challenges and necessitates using advanced models. In our research, conducted in collaboration with Gdańsk University of Technology and the Medical University of Gdańsk, we explore strategies to tackle these challenges. We evaluated the performance of various tools and models in recognizing medical terms within a comprehensive vocabulary, comparing these tools’ outcomes with annotations made by medical experts. Our study specifically examined categories such as ‘Drugs’, ‘Diseases and Symptoms’, ‘Procedures’, and ‘Other Medical Terms’, contrasting human expert annotations with the performance of popular multilingual chatbots and natural language processing (NLP) tools on translated texts. The conclusion drawn from our statistical analysis reveals that no significant differences were detected between the groups we examined. This suggests that the tools and models we tested are, on average, similarly effective—or ineffective—at recognizing medical terms as categorized by our specific criteria. Our findings highlight the challenges in bridging the gap between human and machine accuracy in medical text annotation, especially in non-English contexts, and emphasize the need for further refinement of these technologies.

Pre-trained models have become widely adopted for their strong zero-shot performance, often minimizing the need for task-specific data. However, specialized domains like medical speech recognition still benefit from tailored datasets. We present ADMEDVOICE, a novel Polish medical speech dataset, collected using a high-quality text corpus and diverse recording conditions to reflect real-world scenarios. The dataset includes domain-specific vocabulary such as drug names and illnesses, with nearly 15 hours of audio from 28 speakers, including noisy environments. Additionally, we release two enhanced versions: one anonymized for privacy-sensitive use and another synthetic version created via text-to-speech, totaling over 83 hours and nearly 50,000 samples. Evaluating the Whisper model, we observe a 24.03 WER on our test set. Fine-tuning with human recordings reduces WER to 15.47, and incorporating anonymized and synthetic data further lowers it to 13.91. We open-source the dataset, fine-tuned model, and code on Kaggle to support continued research in medical speech recognition.

Background:Inaccurate blood pressure (BP) classification results in inappropriate treatment. We tested whether machine learning (ML), using routine clinical data, can serve as a reliable alternative to ambulatory BP monitoring (ABPM) in classifying BP status. Methods:This study employed a multicentre approach involving 3 derivation cohorts from Glasgow, Gdańsk, and Birmingham, and a fourth independent evaluation cohort. ML models were trained using office BP, ABPM, and clinical, laboratory, and demographic data, collected from patients referred for hypertension assessment. Seven ML algorithms were trained to classify patients into 5 groups, named as follows: Normal/Target; Hypertension-Masked; Normal/Target-White-Coat (WC); Hypertension-WC; and Hypertension. The 10-year cardiovascular outcomes and 27-year all-cause mortality risks were calculated for the ML-derived groups using the Cox proportional hazards model. Results:Overall, extreme gradient boosting (using XGBoost open source software) showed the highest area under the receiver operating characteristic curve of 0.85-0.88 across derivation cohorts, Glasgow (n = 923; 43% female; age 50.7 ± 16.3 years), Gdańsk (n = 709; 46% female; age 54.4 ± 13 years), and Birmingham (n = 1222; 56% female; age 55.7 ± 14 years). But accuracy (0.57-0.72) and F1 (harmonic mean of precision and recall) scores (0.57-0.69) were low across the 3 patient cohorts. The evaluation cohort (n = 6213; 51% female; age 51.2 ± 10.8 years) indicated elevated 10-year risks of composite cardiovascular events in the Normal/Target-WC and the Hypertension-WC groups, with heightened 27-year all-cause mortality observed in all groups, except the Hypertension-Masked group, compared to the Normal/Target group. Conclusions:ML has limited potential in accurate BP classification when ABPM is unavailable. Larger studies including diverse patient groups and different resource settings are warranted.

Mineralocorticoid receptors are expressed in several structures of the central nervous system, and aldosterone levels can be measured in the brain, although in smaller amounts than in plasma. Nevertheless, these amounts appear to be sufficient to elicit substantial clinical effects. Primary aldosteronism, characterized by high levels of plasma aldosterone, is one of the most common causes of secondary hypertension. In this context, high aldosterone levels may have both indirect and direct effects on the brain with a negative impact on several cerebral functions. Thus, chronic aldosterone excess has been associated with symptoms of anxiety and depression – two clinical entities themselves associated with cognitive deficits. Today, there is an increasing number of reports on the influence of aldosterone on the brain, but there is also a significant amount of uncertainty, such as the role of high aldosterone levels on cognitive functions and decline independently of blood pressure. In this mini review, we discuss the known and unknowns of the impact of aldosterone on the brain putting emphasis on cognitive functions.

We introduce an entropy-based classification method for pairs of sequences (ECPS) for quantifying mutual dependencies in heart rate and beat-to-beat blood pressure recordings. The purpose of the method is to build a classifier for data in which each item consists of two intertwined data series taken for each subject. The method is based on ordinal patterns and uses entropy-like indices. Machine learning is used to select a subset of indices most suitable for our classification problem in order to build an optimal yet simple model for distinguishing between patients suffering from obstructive sleep apnea and a control group.

B a c k g r o u n d: Arterial hypertension (HTN) ranks among the most widespread chronic illnesses that affect adults in industrialized societies.The main goal of this study was to describe the control (inhibition) processes among HTN patients, and to evaluate the dynamics of brain activity while the patients were engaged in tasks measuring the cognitive aspect of self-control.(California Verbal Learning Test, Color Trails Test, The Trail Making Test, Controlled Oral Word Association Test), and a fMRI Stroop test (rapid event design) were administered to 40 persons (20 HTN patients and 20 controls).Groups were matched in terms of age, sex, education, smoking history, and waist-to-hip ratio.R e s u l t s: As revealed by fMRI, the HTN patients demonstrate left-hemisphere asymmetry in inhibitory processes.Also around 90% of patients had problems when completing tasks which rely on verbal and graphomotor aspects of self-control.C o n c l u s i o n s: The results suggest that both cerebral hemispheres must interact correctly in order to provide successful executive control.The deficiencies in control and executive functioning, which were observed among the patients, prove that HTN negatively affects brain processes that control one's cognitive activity. P a r t i c i p a n t s a n d p r o c e d u r e: A set of neuropsychological tests

An analysis of exhaled breath enables specialists to noninvasively monitor biochemical processes and to determine any pathological state in the human body. Breath analysis holds the greatest potential to remold and personalize diagnostics; however, it requires a multidisciplinary approach and collaboration of many specialists. Despite the fact that breath is considered to be a less complex matrix than blood, it is not commonly used as a diagnostic and prognostic tool for early detection of disordered conditions due to its problematic sampling, analysis, and storage. This review is intended to determine, standardize, and marshal experimental strategies for successful, reliable, and especially, reproducible breath analysis.

Subjects with desaturations and sleep apnea episodes during the night characterize with higher variability of parameters describing breathing pattern during wakefulness which might suggest altered respiratory regulation even in mild sleep apnea.

This paper describes a novel way to measure, process, analyze, and compare respiratory signals acquired by two types of devices: a wearable sensorized belt and a microwave radar-based sensor. Both devices provide breathing rate readouts. First, the background research is presented. Then, the underlying principles and working parameters of the microwave radar-based sensor, a contactless device for monitoring breathing, are described. The breathing rate measurement protocol is then presented, and the proposed algorithm for octave error elimination is introduced. Details are provided about the data processing phase; specifically, the management of signals acquired from two devices with different working principles and how they are resampled with a common processing sample rate. This is followed by an analysis of respiratory signals experimentally acquired by the belt and microwave radar-based sensors. The analysis outcomes were checked using Levene's test, the Kruskal-Wallis test, and Dunn's post hoc test. The findings show that the proposed assessment method is statistically stable. The source of variability lies in the person-triggered breathing patterns rather than the working principles of the devices used. Finally, conclusions are derived, and future work is outlined.

Purpose Primary aldosteronism is one of the most frequent causes of secondary arterial hypertension, and whether primary aldosteronism is associated with masked hypertension is unknown. Materials and methods We describe a 64-year-old man with a history of hypothyroidism, recurring hypokalaemia, and normal home and office blood pressure values. Ambulatory blood pressure monitoring revealed masked hypertension with strikingly high systolic blood pressure variability and typical hypertension-mediated organ damage. Results The patient required gradual escalation of antihypertensive medication to four drugs. During the diagnostic process we identified primary aldosteronism, cobalamin deficiency, severe obstructive sleep apnoea, and low baroreflex sensitivity (1.63 ms/mmHg). Following unilateral adrenalectomy, cobalamin supplementation and continuous positive airway pressure, we observed a spectacular improvement in the patient's blood pressure control, baroreflex sensitivity (4.82 ms/mmHg) and quality of life. Conclusions We report an unusual case of both masked arterial hypertension and primary aldosteronism. Elevated blood pressure values were masked in home and office measurements by coexisting hypotension which resulted most probably from deteriorated baroreflex sensitivity. Baroreflex sensitivity increased following treatment, including unilateral adrenalectomy. Hypertension can be masked by coexisting baroreceptor dysfunction which may derive from structural but also functional reversible changes.

Background: Hypertension is a global public health problem. Inaccurate blood pressure (BP) measurements result in unnecessary or insufficient treatment, which increases the risk of adverse drug effects and the cost of healthcare. Ambulatory BP monitoring (ABPM) is the gold standard for assessing hypertension, but costs and patient tolerability limit its use. Using routine clinical data, we aimed to demonstrate the clinical utility of ML-based classification of patients into BP risk groups that are as informative as classifying with an ABPM.Methods: Using office BP and ABPM as well as laboratory, clinical, and demographic data, seven machine learning (ML) algorithms were trained to classify patients referred for hypertension assessment from three cohorts into five groups: Normal/Target, Hypertension-Masked, Normal/Target-White-Coat, Hypertension-White-Coat, and Hypertension. In a fourth independent evaluation cohort, the Cox proportional hazards model was used to calculate the 10-year cardiovascular outcomes and 27-year all-cause mortality risk for the ML-inferred groups.Results: The three derivation cohorts were Glasgow (n=923; 43% females; age 50·7±16·3 years), Gdańsk (n=709; 46% females; age 54·4±13 years), and Birmingham (n=1,222; 56% females; age 55·7±14 years). The model that performed the best was XGBoost (AUROC 0.85-0.88). In the evaluation cohort (n=6213, 51% females; age 51·2±10·8 years), as compared to the referent Normal/Target group, Normal/Target-White-Coat and Hypertension-White-Coat groups had a higher 10-year risk of composite cardiovascular events and all the BP groups except Hypertension-Masked were associated with higher 27-year all-cause mortality.Conclusions: We demonstrate that ML inference of ABPM status from routine clinical data identifies high-risk patient groups for mortality and cardiovascular outcomes. This will improve clinical practice and decrease the burden of hypertension by targeting the use of ABPM where it will be most beneficial, minimizing the burden on patients, and lowering healthcare costs. This will be of particular value in settings with limited resources.